Most conversation dashboards measure activity rather than outcomes. Messages sent, conversations handled, average handle time. All easy to collect, all trivially gameable, and none reliably connected to whether the business is doing better.
A smaller set genuinely predicts revenue. They are harder to instrument, which is precisely why they are worth the effort.
Key takeaways
- Resolution rate matters, but only alongside repeat contact rate.
- CSAT measured only at closure surveys a biased sample.
- Escalation quality predicts churn better than escalation volume.
- Segment every metric by channel. Aggregates conceal the problem.
Metrics that mislead
Deflection rate
Usually defined as the share of conversations closed without a human. It rises when automation improves, and it also rises when customers give up. Those look identical in the data and mean opposite things.
Always pair it with repeat contact rate within 48 hours. High deflection with low repeat contact is genuine success. High deflection with high repeat contact means customers are bouncing off a wall and trying again.
Average handle time
Lower is better only up to a point. Push it as a target and agents close conversations prematurely, which shifts work into tomorrow's queue as repeat contacts. It is a capacity-planning input, not a performance target.
CSAT at closure
Response rates run low, and the people who answer skew to the delighted and the furious. The quietly dissatisfied majority, the ones who churn, do not respond at all.
Every metric that can be improved by doing worse work will eventually be improved by doing worse work.
Metrics worth instrumenting
| Metric | Definition | Why it predicts revenue |
|---|---|---|
| Repeat contact rate | Same customer, same issue, within 48h | Directly measures whether you actually resolved it |
| First response time (p50 and p90) | To first useful answer | Determines whether the conversation continues at all |
| Escalation quality | Share of handoffs needing no repeated questions | Poor handoffs are what customers describe when they churn |
| Conversation-to-outcome rate | Share reaching a defined outcome | Connects support activity to revenue |
| Time to resolution (p90) | Open to genuinely resolved | The tail is where churn originates |
The first is the most underrated in the industry. It cannot be gamed by closing conversations faster, because premature closure increases it. It is the closest thing to an honest quality signal available.
Segment by channel
Aggregating across channels is how real problems disappear. Email and WhatsApp carry expectations an order of magnitude apart; averaging them produces a number describing neither.
Segment every metric by channel, and then by hour of day. A team with excellent daytime performance and a nine-hour overnight gap shows a mediocre average that suggests nothing actionable. Split it and the actual problem is obvious in seconds.
The most useful chart. Median first response time by hour of day, one line per channel. It takes minutes to build and usually reveals the single biggest fixable gap in the operation.
Connecting conversations to revenue
The gap in most reporting is between conversation and outcome. Support reports on conversations, finance reports on revenue, and nothing links them.
Define an outcome per conversation type before instrumenting anything: a booked meeting for pre-sales, a completed order for cart recovery, a retained subscription for cancellations. Then measure the share reaching it.
This reframes the operation from cost centre to revenue contributor, and it changes which improvements get funded. A team that can show cart-recovery conversations converting at a measurable rate is arguing from evidence rather than sentiment.
Frequently asked questions
How many metrics should we track?
Five or six reviewed weekly beats twenty nobody opens. Add one only when you know what decision it changes.
Is CSAT worthless?
No, but it is a lagging indicator with a biased sample. Use it as a trend, not as a target, and never as an individual performance measure.
What is a good repeat contact rate?
It varies by sector, but the trend matters more than the absolute. Rising repeat contact alongside rising deflection is the warning sign to act on.